Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a partially deterministic sampling scheme for compressed sensing using unitary matrices, optimizing Bernoulli selectors to combine random and deterministic row selection. The approach addresses the practical need to incorporate crucial, non-random sampling vectors, which are often ignored in theoretical treatments of compressed sensing. The proposed scheme demonstrably improves image compressed sensing performance with both generative and sparse priors, while also providing improved sample complexity bounds and novel denoising guarantees compared to existing methods.
Rigorously decide which rows to sample deterministically in compressed sensing and get better image recovery with provable guarantees.
We study compressed sensing when the sampling vectors are chosen from the rows of a unitary matrix. In the literature, these sampling vectors are typically chosen randomly; the use of randomness has enabled major empirical and theoretical advances in the field. However, in practice there are often certain crucial sampling vectors, in which case practitioners will depart from the theory and sample such rows deterministically. In this work, we derive an optimized sampling scheme for Bernoulli selectors which naturally combines random and deterministic selection of rows, thus rigorously deciding which rows should be sampled deterministically. This sampling scheme provides measurable improvements in image compressed sensing for both generative and sparse priors when compared to with-replacement and without-replacement sampling schemes, as we show with theoretical results and numerical experiments. Additionally, our theoretical guarantees feature improved sample complexity bounds compared to previous works, and novel denoising guarantees in this setting.